Multi-omics and AI-guided discovery and validation of sisomicin as a COPA-targeting agent for pathological scar inhibition.

Abstract

Pathological scars, including hypertrophic scars and keloids, are fibrotic skin disorders marked by excessive collagen deposition and persistent inflammation, yet effective treatments remain limited. We integrated single-cell RNA sequencing, bulk transcriptomic datasets, and artificial intelligence (AI)-driven machine learning and deep learning to identify molecular drivers of scarring, and subsequently validated our findings using both cellular and animal models. High-dimensional weighted gene co-expression network analysis (hdWGCNA) of scar-associated macrophage subsets revealed COPI coat complex alpha subunit (COPA) as a hub gene correlated with macrophage and fibroblast infiltration. AI-guided drug screening and molecular docking identified sisomicin, an aminoglycoside antibiotic, as a potential COPA-targeting agent. Functional assays in human hypertrophic scar-derived fibroblasts (HHFs) and TGF-β1-stimulated NFs showed that sisomicin inhibited proliferation, migration, and collagen synthesis, while increasing ROS and apoptosis. In bleomycin- and traction-induced mouse models, sisomicin reduced dermal thickening, collagen deposition, and COPA expression. This multi-omics and AI-based framework uncovers COPA as a key regulator of pathological scars and highlights sisomicin as a promising therapeutic candidate, offering a broadly applicable strategy for anti-fibrotic drug discovery.